<LINK "wei-n*">"wei-r4"> <TARGET "wei" DOCINFO AUTHOR "Jeroen van de Weijer" TITLE "An Optimality Theoretical analysis of the Dutch diminutive" SUBJECT "AVT Publications 19" KEYWORDS "" SIZE HEIGHT "220" WIDTH "150" VOFFSET "4"> An Optimality Theoretical analysis of the Dutch diminutive* Jeroen van de Weijer Leiden University/ULCL 1. Introduction The analysis of the diminutive in Dutch has been a topic of debate among phonolo- gists and morphologists for almost one and a half centuries.1 There are five allo- morphs in the standard language, there is a lot of variation between dialects, and there is variation between speakers for diminutives of morphologically complex and rarely used words. As a result, previous analyses have argued over the data, the shape of the underlying form of the morpheme and the sets of rules which would derive the surface alternants. These rules have usually been highly complex, and idiosyncratic for the diminutive — that is, the phonological rules concerned were marked for applying only in the context of the diminutive morpheme. On a theoretical level, morpheme-specific rules are a problem in a framework such as Optimality Theory, in which constraints are assumed to be universal, i.e. not specific to particular languages and, a fortiori, not specific to particular morphemes in particular languages. Thus, the Dutch diminutive, with at least three such morphophonological rules, is an obvious challenge to the OT framework and invites re-analysis. In this investigation, I present the outlines of an OT-style analysis of the allomorphy in the standard language, comparing this to earlier rule-based work. I will show that the language-specificness can be circumvented to a large degree in a constraint-based framework, which, additionally, captures a number of phonologi- cal regularities that are not expressed by the rule-based approach. 2. Data and rule-based analysis In this section I present the basic alternants of the diminutive in standard Dutch, and review one previous analysis, presented in Gussenhoven & Jacobs (1998). Other Linguistics in the Netherlands 2002, 199–209. issn 0929–7332 / e-issn 1569–9919©John Benjamins Publishing Company <LINK "wei-r9">"wei-r15">"wei-r7">"wei-r4"> 200 Jeroen van de Weijer work includes Kooij (1982), Trommelen (1983), and van der Hulst (1984), among many others. The basic facts of the diminutive are presented in (1) (data partly taken from van der Hulst 1984:117ff., where more examples can be found). (1) a. -6tj6 after short vowel + liquid or nasal: karretje ‘car-dim’ balletje ‘ball-dim’ b. -j6 after obstruent: busje ‘bus-dim’ kerkje ‘church-dim’ c. -pj6 after long vowel + m: duimpje ‘thumb-dim’ raampje ‘window-dim’ after short vowel + liquid + m: halmpje ‘stalk-dim’ bermpje ‘verge-dim’ d. -kj6 after short vowel + ]: koninkje ‘king-dim’ landinkje ‘landing-dim’ e. -tj6 after long vowel + son: laantje ‘avenue-dim’ boeitje ‘buoy-dim’ after long vowels and diphthongs: zeetje ‘sea-dim’ uitje ‘onion-dim’ after short vowel+r+n: kerntje ‘core-dim’ after schwa:2 ?wondetje ‘wound-dim’ ?boetetje ‘fine-dim’ ?hetzetje ‘innuendo-dim’ The form /tj6/ has been taken as the underlying form in most previous analyses, because this is the one which appears after vowels, while the other alternants (especially [pj6], [kj6], [j6]) can be derived from it rather straightforwardly. The epenthesised form of the diminutive, [6tj6], has evoked more discussion; from the examples above it is clear that, a priori, the length of the preceding vowel and the nature of the preceding consonant play a role. Gussenhoven & Jacobs (1998) (henceforth GJ) present an account of the diminutive allomorphy based on SPE-type rules. First, assuming underlying /tj6/, a rule is necessary to convert /t/ into [p] or [k] after labials and velars, respectively. Examples of this alternation were given in (1c,d) above. The rule GJ formulate for this (1998:108) is given in (2): È +nas ˘ (2) place assimilation t Æ [αplace] /Í˙ + __ j6]dim Î αplace ˚ Note that this rule must refer specifically to the diminutive, since Place assimilation <LINK "wei-r12"> An Optimality Theoretical analysis of the Dutch diminutive 201 of nasals to other consonants in Dutch is regressive in all cases other than the diminutive, as illustrated by the data in (3): (3) in+brengen ‘to bring in’ /n+b/ Æ [mb] in+kopen ‘to purchase’ /n+k/ Æ []k] in+gaan ‘to enter’ /n+>/ Æ []>] man#kracht ‘manpower’ /n+k/ Æ []k] In the forms in (3) the stop determines the place of articulation of the preceding nasal, while the place of the stop in the diminutive allomorph is determined by the last consonant of the root through progressive assimilation. Another rule which is idiosyncratic for the diminutive is t-deletion, responsi- ble for deletion of /t/ after obstruents. Data were presented in (1b). The rule responsible for this is given below (GJ 1998:108): (4) t-deletion t Æ Ø / [−son] + __ j6]dim Although Dutch has more cases of t-deletion, this phenomenon must be formulated to apply specifically in diminutives, since in compounds, regular inflection and non-derived words, no deletion takes place after single obstruents: (5) huis-tuin… ‘house-garden…’ pas-te ‘fit-past’ woestijn ‘desert’ The most spectacular alternation concerns the insertion of schwa after roots that end in a short vowel followed by a single sonorant. Examples were given in (1a) above. The class of consonants after the short vowel can be identified with the sonorants; practically only nasals and liquids occur in this position, as there are only very few words in Dutch that have short vowel + glide, such as koi ~ koitje ‘Japanese carp’.If the final glide is syllabified in the nucleus (making [fI] a real diphthong, on a par with ui [œy]) the diminutive will be [tj6] here, too. The rule GJ (1998:107) formulate is given in (6), where the short vowels are identified with the class of lax vowels: È +cons ˘ (6) 6-insertion Ø Æ 6 / [−tense]Í˙ + __ tj6]dim Î +son ˚ Again, the rule is specific to the diminutive, since schwa insertion does not take place in similar environments in underived words (e.g. the Indonesian name Kantjil is never pronounced (*[kan6tjil]) cf. [kan6tj6] from /kan+tj6/) or in other types of derivation. To conclude, a rule-based account of the Dutch diminutive needs three rules specific to this morpheme ((2), (4) and (6)) to account for the alternations induced by the diminutive morpheme in Dutch. Optimality Theory (Prince & Smolensky 1993) does not use rules, of course, but only constraints, which are assumed to be <LINK "wei-r10">"wei-r8"> 202 Jeroen van de Weijer universal, not language-specific. Hence, an OT account making use of constraints pertaining only to the diminutive morpheme in Dutch is not adequate. In the next section we will see to what extent an alternative OT account of the data is possible. 3. Towards an OT account First, let us try to deal with Place assimilation [tj6~pj6~kj6] in diminutives. In OT, it is necessary to assume a constraint against heterorganic nasal-stop sequences. One possible formulation is given in (7): (7) NasalPlaceAgreement (NPA) A sequence of nasal plus consonant must be homorganic. There is a large variety of specific proposals that will produce this effect, and we will not explore the specifics here. It is obvious that the constraint in (7) has a wider scope than the rule in (2) that was supposed to account for the same kind of Place alternations. For instance, the constraint in (7) also accounts for the facts in (3), and would therefore alone seem to be preferable to the rule-based account. Note that identity of Place can be achieved in (at least) two ways: by changing the nasal (which would result in the incorrect output form [bo˜ntj6] or by changing the stop consonant in the diminutive, which gives the correct [bo˜mpj6]. It is an important insight that in such cases in Dutch the base is never altered: alterations target the diminutive morpheme, not the root. This is expressed by the following well- established, and tentatively universal, constraint ranking (McCarthy & Prince 1995, see also Kager 1999:75–76): (8) Faith(Root)»Faith(Affix) Faithfulness requirements are enforced more strictly within the root than in non-root morphemes, such as affixes. This accounts for the difference in direction of assimilation between the data in (1) and the facts in (3): the crucial insight is that in both cases the affix is targeted and the root is left intact.3 Note that the fact that Faith(Root) dominates Faith(Affix) does not mean, of course, that the former is undominated: Faith(Root) can be violated to satisfy higher level constraints. For instance, final root consonants can be devoiced, and thus altered, to satisfy whatever constraint configuration is necessary to express final devoicing in Dutch. Tableau (9) illustrates the selection of the output [bo˜mpj6], where both Faith-constraints are cover terms for a family of correspondence constraints, in particular Ident (McCarthy & Prince 1995): An Optimality Theoretical analysis of the Dutch diminutive 203 (9) /bo˜m/+/tj6/ NPA Faith(Rt) Faith(Aff) a. [bo˜mtj6] *! b. [bo˜ntj6] *! c. [bo˜mpj6] * It must be pointed out that NPA is violated in other parts of the morphology of Dutch. Consider the use of the distributive suffix (or rather circumfix) ge- + -te in (10), for instance: (10) ge+boom+te tree-dist ‘foliage’ ge+raam+te window-dist ‘skeleton’ ge+worm+te worm-dist ‘vermin’ Although this suffix is hardly productive, the question might be raised how assimilation is blocked in these cases (preventing *geboompe, on a par with boom- pje).